This invention relates to methods for the preparation and use of variant antibodies and finds application particularly in the fields of immunology and cancer diagnosis and therapy.
Naturally occurring antibodies (immunoglobulins) comprise two heavy chains linked together by disulfide bonds and two light chains, one light chain being linked to each of the heavy chains by disulfide bonds. Each heavy chain has at one end a variable domain (VH) followed by a number of constant domains. Each light chain has a variable domain (VL) at one end and a constant domain at its other end, the constant domain of the light chain is aligned with the first constant domain of the heavy chain, and the light chain variable domain is aligned with the variable domain of the heavy chain. Particular amino acid residues are believed to form an interface between the light and heavy chain variable domains, see e.g. Chothia et al., J. Mol. Biol. 186:651-663 (1985); Novotny and Haber, Proc. Natl. Acad. Sci. USA 82:4592-4596 (1985).
The constant domains are not involved directly in binding the antibody to an antigen, but are involved in various effector functions, such as participation of the antibody in antibody-dependent cellular cytotoxicity. The variable domains of each pair of light and heavy chains are involved directly in binding the antibody to the antigen. The domains of natural light and heavy chains have the same general structure, and each domain comprises four framework (FR) regions, whose sequences are somewhat conserved, connected by three hyper-variable or complementarity determining regions (CDRs) (see Kabat, E. A. et a., Sequences of Proteins of Immunological Interest, National Institutes of Health, Bethesda, Md., (1987)). The four framework regions largely adopt a xcex2-sheet conformation and the CORs form loops connecting, and in some cases forming part of, the xcex2-sheet structure. The CDRs in each chain are held in close proximity by the framework regions and, with the CDRs from the other chain, contribute to the formation of the antigen binding site.
Widespread use has been made of monoclonal antibodies, particularly those derived from rodents including mice, however they are frequently antigenic in human clinical use. For example, a major limitation in the clinical use of rodent monoclonal antibodies is an anti-globulin response during therapy (Miller, R. A. et al., Blood 62:988-995 (1983); Schroff, R. W. et al., Cancer Res. 45:879-885 (1985)).
The art has attempted to overcome this problem by constructing xe2x80x9cchimericxe2x80x9d antibodies in which an animal antigen-binding variable domain is coupled to a human constant domain (Cabilly et al., U.S. Pat. No. 4,816,567; Morrison, S. L. et al., Proc. Natl. Acad. Sci. USA 81:6851-6855 (1984); Boulianne, G. L. et al., Nature 312:643-646 (1984); Neuberger, M. S. et al., Nature 314:268-270 (1985)). The term xe2x80x9cchimericxe2x80x9d antibody is used herein to describe a polypeptide comprising at least the antigen binding portion of an antibody molecule linked to at least part of another protein (typically an immunoglobulin constant domain).
The isotype of the human constant domain may be selected to tailor the chimeric antibody for participation in antibody-dependent cellular cytotoxicity (ADCC) and complement-dependent cytotoxicity (see e.g. Brxc3xcggemann, M. et al., J. Exp. Med. 166:1351-1361 (1987); Riechmann, L. et al., Nature 332:323-327 (1988); Love et al., Methods in Enzymology 178:515-527 (1989); Bindon et al., J. Exp. Med. 168:127-142 (1988).
In the typical embodiment, such chimeric antibodies contain about one third rodent (or other non-human species) sequence and thus are capable of eliciting a significant anti-globulin response in humans. For example, in the case of the murine anti-CD3 antibody, OKT3, much of the resulting anti-globulin response is directed against the variable region rather than the constant region (Jaffers, G. J. et a., Transplantation 41:572-578 (1986)).
In a further effort to resolve the antigen binding functions of antibodies and to minimize the use of heterologous sequences in human antibodies, Winter and colleagues (Jones, P. T. et al., Nature 321:522-525 (1986); Riechmann, L. et al., Nature 332:323-327 (1988); Verhoeyen, M. et al., Science 239:1534-1536 (1988)) have substituted rodent CDRs or CDR sequences for the corresponding segments of a human antibody. As used herein, the term xe2x80x9chumanizedxe2x80x9d antibody is an embodiment of chimeric antibodies wherein substantially less than an intact human variable domain has been substituted by the corresponding sequence from a non-human species. In practice, humanized antibodies are typically human antibodies in which some CDR residues and possibly some FR residues are substituted by residues from analogous sites in rodent antibodies.
The therapeutic promise of this approach is supported by the clinical efficacy of a humanized antibody specific for the CAMPATH-1 antigen with two non-Hodgkin lymphoma patients, one of whom had previously developed an anti-globulin response to the parental rat antibody (Riechmann, L. et al., Nature 332:323-327 (1988); Hale, G. et al., Lancet i:1394-1399 (1988)). A murine antibody to the interleukin 2 receptor has also recently been humanized (Queen, C. et al., Proc. Natl. Acad. Sci. USA 86:10029-10033 (1989)) as a potential immunosuppressive reagent. Additional references related to humanization of antibodies include Co et al., Proc. Natl. Acad. Sci. USA 88:2869-2873 (1991); Gorman et al., Proc. Natl. Acad. Sci. USA 88:4181-4185 (1991); Daugherty et al., Nucleic Acids Research 19(9):2471-2476 (1991); Brown et al., Proc. Natl. Acad. Sci. USA 88:2663-2667 (1991); Junghans et al., Cancer Research 50:1495-1502 (1990).
In some cases, substituting CDRs from rodent antibodies for the human CDRs in human frameworks is sufficient to transfer high antigen binding affinity (Jones, P. T. et al., Nature 321:522-525 (1986); Verhoeyen, M. et al., Science 239:1534-1536 (1988)), whereas in other cases it has been necessary to additionally replace one (Riechmann, L. et al., Nature 332:323-327 (1988)) or several (Queen, C. et al., Proc. Natl. Acad. Sci. USA 86:10029-10033 (1989)) framework region (FR) residues. See also Co et al., supra.
For a given antibody a small number of FR residues are anticipated to be important for antigen binding. Firstly for example, certain antibodies have been shown to contain a few FR residues which directly contact antigen in crystal structures of antibody-antigen complexes (e.g., reviewed in Davies, D. R. et al., Ann. Rev. Biochem. 59:439-473 (1990)). Secondly, a number of FR residues have been proposed by Chothia, Lesk and colleagues (Chothia, C. and Lesk, A. M., J. Mol. Biol. 196:901-917 (1987); Chothia, C. et al., Nature 342:877-883 (1989); Tramontano, A. et al., J. Mol. Biol. 215:175-182 (1990)) as critically affecting the conformation of particular CDRs and thus their contribution to antigen binding. See also Margolies et al., Proc. Natl. Acad. Sci. USA 72:2180-2184 (1975).
It is also known that, in a few instances, an antibody variable domain (either VH or VL) may contain glycosylation sites, and that this glycosylation may improve or abolish antigen binding, Pluckthun, Biotechnology 9:545-51 (1991); Spiegelberg et al., Biochemistry 9:4217-4223 (1970); Wallic et al., J. Exp. Med. 168:1099-1109 (1988); Sox et al., Proc. Natl. Acad. Sci. USA 66:975-982 (1970); Margni et al., Ann. Rev. Immunol 6:535-554 (1988). Ordinarily, however, glycosylation has no influence on the antigen-binding properties of an antibody, Pluckthun, supra, (1991).
The three-dimensional structure of immunoglobulin chains has been studied, and crystal structures for intact immunoglobulins, for a variety of immunoglobulin fragments, and for antibody-antigen complexes have been published (see e.g., Saul et al., Journal of Biological Chemistry 25:585-97 (1978); Sheriff et al., Proc. Natl. Acad. Sci. USA 84:8075-79 (1987); Segal et al., Proc. Natl. Acad. Sci. USA 71:4298-4302 (1974); Epp et al., Biochemistry 14(22):4943-4952 (1975); Marquart et al., J. Mol. Biol. 141:369-391 (1980); Furey et al., J. Mol. Biol. 167:661-692 (1983); Snow and Amzel, Protein: Structure, Function, and Genetics 1:267-279, Alan R. Liss, Inc. pubs. (1986); Chothia and Lesk, J. Mol. Bol. 196:901-917 (1987); Chothia et al., Nature 342:877-883 (1989); Chothia et al., Science 233:755-58 (1986); Huber et al., Nature 264:415-420 (1976); Bruccoleri et al., Nature 335:564-568 (1988) and Nature 336:266 (1988); Sherman et al., Journal of Biological Chemistry 263:4064-4074 (1988); Amzel and Poljak, Ann. Rev. Biochem. 48:961-67 (1979); Silverton et al., Proc. Natl. Acad. Sci. USA 74:5140-5144 (1977); and Gregory et al., Molecular Immunology 24:821-829 (1987). It is known that the function of an antibody is dependent on its three dimensional structure, and that amino acid substitutions can change the three-dimensional structure of an antibody, Snow and Amzel, supra. It has previously been shown that the antigen binding affinity of a humanized antibody can be increased by mutagenesis based upon molecular modelling (Riechmann, L. et al., Nature 332:323-327 (1988); Queen, C. et al., Proc. Natl. Acad. Sci. USA 86:10029-10033 (1989)).
Humanizing an antibody with retention of high affinity for antigen and other desired biological activities is at present difficult to achieve using currently available procedures. Methods are needed for rationalizing the selection of sites for substitution in preparing such antibodies and thereby increasing the efficiency of antibody humanization.
The proto-oncogene HER2 (human epidermal growth factor receptor 2) encodes a protein tyrosine kinase (p185HER2) that is related to and somewhat homologous to the human epidermal growth factor receptor (see Coussens, L. et al., Science 230:1132-1139 (1985); Yamamoto, T. et al., Nature 319:230-234 (1986); King, C. R. et al., Science 229:974-976 (1985)). HER2 is also known in the field as c-erbB-2, and sometimes by the name of the rat homolog, neu. Amplification and/or overexpression of HER2 is associated with multiple human malignancies and appears to be integrally involved in progression of 25-30% of human breast and ovarian cancers (Slamon, D. J. et al., Science 235:177-182 (1987), Slamon, D. J. et al., Science 244:707-712 (1989)). Furthermore, the extent of amplification is inversely correlated with the observed median patient survival time (Slamon, supra, Science 1989).
The murine monoclonal antibody known as muMAb4D5 (Fendly, B. M. et al., Cancer Res. 50:1550-1558 (1990)), directed against the extracellular domain (ECD) of p185HER2, specifically inhibits the growth of tumor cell lines overexpressing p185HER2 in monolayer culture or in soft agar (Hudziak, R. M. et al., Molec. Cell. Biol 9:1165-1172 (1989); Lupu, R. et al., Science 249:1552-1555 (1990)). MuMAb4D5 also has the potential of enhancing tumor cell sensitivity to tumor necrosis factor, an important effector molecule in macrophage-mediated tumor cell cytotoxicity (Hudziak, supra, 1989; Shepard, H. M. and Lewis, G. D. J. Clinical Immunology 8:333-395 (1988)). Thus muMAb4D5 has potential for clinical intervention in and imaging of carcinomas in which p185HER2 is overexpressed. The muMAb4D5 and its uses are described in PCT application WO 89/06692 published Jul. 27, 1989. This murine antibody was deposited with the ATCC and designated ATCC CRL 10463. However, this antibody may be immunogenic in humans.
It is therefore an object of this invention to provide methods for the preparation of antibodies which are less antigenic in humans than non-human antibodies but have desired antigen binding and other characteristics and activities.
It is a further object of this invention to provide methods for the efficient humanization is of antibodies, i.e. selecting non-human amino acid residues for importation into a human antibody background sequence in such a fashion as to retain or improve the affinity of the non-human donor antibody for a given antigen.
It is another object of this invention to provide humanized antibodies capable of binding p185HER2.
Other objects, features, and characteristics of the present invention will become apparent upon consideration of the following description and the appended claims.
The objects of this invention are accomplished by a method for making a humanized antibody comprising amino acid sequence of an import, non-human antibody and a human antibody, comprising the steps of:
a. obtaining the amino acid sequences of at least a portion of an import antibody variable domain and of a consensus variable domain;
b. identifying Complementarity Determining Region (CDR) amino acid sequences in the import and the human variable domain sequences;
c. substituting an import CDR amino acid sequence for the corresponding human CDR amino acid sequence;
d. aligning the amino acid sequences of a Framework Region (FR) of the import antibody and the corresponding FR of the consensus antibody;
e. identifying import antibody FR residues in the aligned FR sequences that are non-homologous to the corresponding consensus antibody residues;
f. determining if the non-homologous import amino acid residue is reasonably expected to have at least one of the following effects:
1. non-covalently binds antigen directly,
2. interacts with a CDR; or
3. participates in the VL-VH interface; and
g. for any non-homologous import antibody amino acid residue which is reasonably expected to have at least one of these effects, substituting that residue for the corresponding amino acid residue in the consensus antibody FR sequence.
Optionally, the method of this invention comprises the additional steps of determining if any non-homologous residues identified in step (e) are exposed on the surface of the domain or buried within it, and if the residue is exposed but has none of the effects identified in step (f), retaining the consensus residue.
Additionally, in certain embodiments the method of this invention comprises the feature wherein the corresponding consensus antibody residues identified in step (e) above are selected from the group consisting of 4L, 35L, 36L, 38L, 43L, 44L, 46L, 58L, 62L, 63L, 64L, 65L, 66L, 67L, 68L, 69L, 70L, 71 L, 73L, 85L, 87L, 98L, 2H, 4H, 24H, 36H, 37H, 39H, 43H, 45H, 49H, 58H, 60H, 67H, 68H, 69H, 70H, 73H, 74H, 75H, 76H, 78H, 91H, 92H, 93H, and 103H (utilizing the numbering system set forth in Kabat, E. A. et al., Sequences of Proteins of Immunological Interest (National Institutes of Health, Bethesda, Md., 1987)).
In certain embodiments, the method of this invention comprises the additional steps of searching either or both of the import, non-human and the consensus variable domain sequences for glycosylation sites, determining if the glycosylation is reasonably expected to be important for the desired antigen binding and biological activity of the antibody (i.e., determining if the glycosylation site binds to antigen or changes a side chain of an amino acid residue that binds to antigen, or if the glycosylation enhances or weakens antigen binding, or is important for maintaining antibody affinity). If the import sequence bears the glycosylation site, it is preferred to substitute that site for the corresponding residues in the consensus human if the glycosylation site is reasonably expected to be important. If only the consensus sequence, and not the import, bears the glycosylation site, it is preferred to eliminate that glycosylation site or substitute therefor the corresponding amino acid residues from the import sequence.
Another embodiment of this invention comprises aligning import antibody and the consensus antibody FR sequences, identifying import antibody FR residues which are non-homologous with the aligned consensus FR sequence, and for each such non-homologous import antibody FR residue, determining if the corresponding consensus antibody residue represents a residue which is highly conserved across all species at that site, and if it is so conserved, preparing a humanized antibody which comprises the consensus antibody amino acid residue at that site.
Certain alternate embodiments of the methods of this invention comprise obtaining the amino acid sequence of at least a portion of an import, non-human antibody variable domain having a CDR and a FR, obtaining the amino acid sequence of at least a portion of a consensus antibody variable domain having a CDR and a FR, substituting the non-human CDR for the human CDR in the consensus antibody variable domain, and then substituting an amino acid residue for the consensus amino acid residue at at least one of the following sites:
a. (in the FR of the variable domain of the light chain) 4L, 35L, 36L, 38L, 43L, 44L, 58L, 46L, 62L, 63L, 64L, 65L, 66L, 67L, 68L, 69L, 70L, 71L, 73L, 85L, 87L, 98L, or
b. (in the FR of the variable domain of the heavy chain) 2H, 4H, 24H, 36H, 37H, 39H, 43H, 45H, 49H, 58H, 60H, 67H, 68H, 69H, 70H, 73H, 74H, 75H, 78H, 91H, 92H, 93H, and 103H.
In preferred embodiments, the non-CDR residue substituted at the consensus FR site is the residue found at the corresponding location of the non-human antibody.
Optionally, this just-recited embodiment comprises the additional steps of following the method steps appearing at the beginning of this summary and determining whether a particular amino acid residue can reasonably be expected to have undesirable effects.
This invention also relates to a humanized antibody comprising the CDR sequence of an import, non-human antibody and the FR sequence of a human antibody, wherein an amino acid residue within the human FR sequence located at any one of the sites 4L, 35L, 36L, 38L, 43L, 44L, 46L, 58L, 62L, 63L, 64L, 65L, 66L, 67L, 68L, 69L, 70L, 2H, 4H, 24H, 36H, 37H, 39H, 43H, 45H, 49H, 58H, 60H, 67H, 68H, 69H, 70H, 73H, 74H, 75H, 76H, 78H, 91H, 92H, 93H, and 103H has been substituted by another residue. In preferred embodiments, the residue substituted at the human FR site is the residue found at the corresponding location of the non-human antibody from which the non-human CDR was obtained. In other embodiments, no human FR residue other than those set forth in this group has been substituted.
This invention also encompasses specific humanized antibody variable domains, and isolated polypeptides having homology with the following sequences.
1. SEQ. ID NO. 1, which is the light chain variable domain of a humanized version of muMAb4D5: DIQMTOSPSSLSASVGDRVTITCRASQDVNTAVAWYQQKPGKAPKLLIYSASFLESGVPSRFSGSRSGTDFTLTISSLQPEDFATYYCQQHYTTPPTFGQGTKVEIKRT
2. SEQ. ID NO. 2, which is the heavy chain variable domain of a humanized version of muMAb4D5): EVQLVESGGGLVOPGGSLRLSCAASGFNIKDTYIHWVRQAPGKGLEWVARIYPTNGYTRYADSVKGRFTISADTSKNTAYLQMNSLRAEDTAVYYCSRWGGDGFYAMDVWGQGTLVTVSS
In another aspect, this invention provides a consensus antibody variable domain amino acid sequence for use in the preparation of humanized antibodies, methods for obtaining, using, and storing a computer representation of such a consensus sequence, and computers comprising the sequence data of such a sequence. In one embodiment, the following consensus antibody variable domain amino acid sequences are provided:
SEQ. ID NO. 3 (light chain): DDIOMTQSPSSLSASVGDRVTITCRASQDVSSYLAWYQQKPGKAPKLLIYAASSLESGVPSRFSGSGSGTDFTLTISSLQPEDFATYYCQQYNSLPYTFGQGTKVEIKRT, and
SEQ. ID NO. 4 (heavy chain): EVQLVESGGGLVQPGGSLRLSCAASGFTFSDYAMSWVRQAPGKGLEWVAVISENGGYTRYADSVKGRFTISADTSKNTAYLQMNSLRAEDTAWYCSRWGGDGFYAMDVWGQGTLVTVSS